INFECTION AND IMMUNITY, Apr. 1974, p. 719-729Copyright 0 1974 American Society for Microbiology

Vol. 9, No. 4Printed in U.S.A.

Predominant Gram-Positive Bacteria in Human Feces:Numbers, Variety, and Persistence

JENNIFER GOSSLING AND JOHN M. SLACK.

Department of Microbiology, West Virginia University Medical Center, Morgantown, West Virginia 26506

Received for publication 15 October 1973

The predominant gram-positive bacteria in 47 fecal specimens from 10 healthymen were studied by microscopic and cultural counts, by the characterizationand tentative identification of isolates, and by the use of fluorescein isothiocya-nate (FITC)-conjugated globulins prepared using some of the isolates. Gram-positive bacteria averaged 101o.5140 (SD/g (wet weight) of feces with significantvariation from host to host. Characterization of 865 isolates, all strict anaerobesand carbohydrate fermenters, showed 12 to 39 distinguishable strainas from eachhost and indicated that some strains were present the full period of about 18months. Sixty percent of the isolates belonged to one of five types, tentativelyidentified with five species-Bifidobacterium adolescentis, Eubacteriumaerofaciens, E. rectale, Peptostreptococcus productus, and Ruminococcus bro-mii. There was distinct host idiosyncrasy in the pattern of estimated counts ofthese five types. Certain strains resembling B. adolescentis, E. aerofaciens, andP. productus, distinguished with FITC conjugates, were resident in their hosts formany months. In direct smears each strain constituted about 1% of the totalbacteria.

There have been many surveys of the humanfecal flora which have shown that anaerobicgram-positive bacteria are very numerous,numbering more than 10'/g (wet weight) offeces (5, 9, 14, 15, 20; S. M. Finegold and L. G.Miller, Bacteriol. Proc., p. 93, 1968). In certainof the surveys some or all of these bacteria wereshown to be acid tolerant and were describedsimply as bifidobacteria or anaerobic lactoba-cilli. Other investigators have reported a varietyof gram-positive bacteria isolated in large num-bers from human feces and have identified themas species of Bifidobacterium, Eubacterium,Peptococcus, Peptostreptococcus, Propionibac-terium, and Ruminococcus (1, 11, 16).Some of the surveys showed that, for certain

types of bacteria at least, there is a tendency tohost idiosyncrasy in the composition of the fecalflora which persists over a period of weeks ormonths (7, 24). Other investigations have indi-cated stability by demonstrating that particularstrains of Escherichia coli may persist in thehuman intestine for long periods (2, 6, 19, 23).The major purpose of this investigation was

to characterize the predominant gram-positivebacteria in fecal samples from 10 hosts and todetermine the extent of quantitative and quali-tative variation both between hosts and withinhosts over a period of more than a year.

MATERIALS AND METHODSSubjects. Ten healthy men, 20 to 40 years old,

consuming an average American diet, were studiedover a period of about 18 months. Two had hadappendectomies, and the remainder had no history ofmajor gastrointestinal tract disorder. No specimenswere collected until at least 3 weeks after any antimi-crobial treatment or diarrheal disease.

Media. The ingredients for the rumen fluid agar(RF), peptone yeast sugar (PYS) agar and broth,fermentation end-products broth (FEP), dilutingfluid (DF), antigen broth (AB), and fermentation testmedium (FTM) are given in Table 1. The first fivemedia were prereduced and anaerobically sterilized(11). Before autoclaving they were dispensed intohard glass tubes or flasks which were sealed withnotched black rubber stoppers; whenever a tube orflask was opened, a stream of sterile oxygen-free gas(11) was used to maintain anaerobic conditions. PYSagar and FTM were prepared as agar deeps by mixingand boiling the ingredients, adding cysteine, and thendistributing and autoclaving in the conventionalmanner. They were used immediately or reheated justbefore use.

Processing specimens. A 1-g sample was takenfrom the center of the fecal specimen within 15 min ofpassage and was suspended in 9 ml of DF. Anadditional 1-g sample was weighed in a tared pan fordry weight estimation. Serial 10-fold dilutions in DFwere prepared from the suspension. Two tubes ofmolten RF agar were inoculated with 0.1 ml from eachof the 10', 10-', and 10-9 dilutions. The tubes were

719

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GOSSLING AND SLACK

TABLE 1. Composition of media

Quantity used in 1 liter of medium

Medium - c E _ -aPYS~ .-. a20 1

FEP~~~~~~~~~~~5.05.0..)05l04 :0C2

0 22CZ0

RF 0.25 0.25 0.5 20 0.5 300 0.5 1.0 1.0 500 200 Co2PYS 2.5 2.5 20 10 (20)' 0.5 2.0 0.5 1.0 500 500 CO2FEP 5.0 5.0 5.0 5.0 0.5 1.0 40 960 C02:N2

1:9DF 0.5 0.5 500 500 Co2AB 1.0 1.0 10 5.0 2.0 0.5 1.0 40 960 C02: N2

1:9FTM - jib1 5.0 5.0 J0.5 40 40 960 Air

a PYS was prepared either with 20 g of agar for roll tubes or deeps or without agar as a broth.

rolled while cooling (11) and incubated for 5 days at39 C. Slides were prepared from the 10-' and 10'dilutions by spreading 0.01 ml over 1 cm2 and heatfixing. Two smears were prepared separately on eachslide. The slides were stored in the refrigerator andagain heat fixed before staining.

Counts. One slide with two smears of the 10-4dilution was stained with freshly filtered rose bengal(rose bengal, 1 g; phenol, 5 g; calcium chloride, 0.02 g;in 100 ml of distilled water) for 3 min. The individualbacteria were counted in 50 fields of each smear at amagnification of x950. If the counts for the twosmears differed by more than 5% of the total, thecount was rejected and another slide was stained andcounted. Allowing for the dilutions and the size of thefield, the bacteria counted were multiplied by 8.414 x107 to give the number of bacteria per gram (wetweight). Another slide was Gram stained. Gram-nega-tive and -positive rods and cocci were counted in 25fields in each smear. The ratios of these types werethen applied to the total number of bacteria per gram(wet weight) obtained above to estimate the numbersof these different types per gram (wet weight).Rumen fluid agar roll tubes with 50 to 500 colonies

were selected for colony counts. The counts from tworeplicate cultures were averaged and multiplied bythe dilution factors to give counts per gram (wetweight). Fifty unselected colonies were Gram stainedand classified as gram positive or gram negative andas rods or cocci. The ratios of the different types werethen applied to the total cultural count to giveestimates of the cultural counts of these differenttypes per gram (wet weight).The 1-g sample in the tared pan was dried to

constant weight in an evacuated desiccator overanhydrous calcium sulfate. The dry weight of nonmi-crobial residue was estimated by subtracting anestimate of the dry weight of the bacteria (directcount/gram [wet weight] x 2 x 10-13 [reference 17])from the total dry weight. Counts per gram (dry

weight) and per gram (dry residue) were obtained bydividing the counts per gram (wet weight) by the dryweight and the dry residue, respectively.

All arithmetical and statistical calculations on theraw data from these counts were performed on theWest Virginia University IBM 360/75 computer usingthe Statistical Analysis System (21).

Isolation and characterization of bacteria. Colo-nies, usually 60, were picked and streaked on rolledPYS agar which was incubated at 35 C. All transferswere made using anaerobic precautions (11). The daysrequired for visible growth on the initial PYS subcul-ture were recorded, and then the colonies were Gramstained. Colonies of gram-positive bacteria were re-streaked on PYS agar until pure cultures were ob-tained, and then a PYS broth was inoculated. Thisbroth culture was examined by Gram stain, and someof it was lyophilized or stored frozen at -60 C. A PYSbroth culture was used to inoculate a PYS agar deep,an FEP broth, and one tube each of FTM withglucose, maltose, lactose, sucrose, soluble starch,mannose, xylose, inulin, mannitol, or inositol; FTMwith amygdalin was also used for suspected Eubacter-ium aerofaciens.The PYS agar deeps were incubated for 3 days and

examined for depth of growth, reduction of theindicator, and production of acid and gas. The FTMcultures were examined after 2 and 7 days of incuba-tion for acid and gas production. To determine theend products of fermentation, the FEP broth wasincubated for 2 days and centrifuged at 800 x g for 30min, and the supernatant fluid was acidified (11) andstored sealed at -60 C until it was examined for thepresence of volatile fatty acids, alcohols, lactate, andsuccinate by gas-liquid chromatography. Three mi-croliters of this fluid was injected into a glass columncontaining acid-washed Chromosorb W, 80/100 mesh2coated with SP1200/phosphoric acid, 10/1% (Su-pelco). The injection chamber was at 130 C, and thecolumn was at a constant 120 C with nitrogen flowing

720 INFECT. IMMUNITY

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GRAM-POSITIVE BACTERIA IN HUMAN FECES

at 40 ml/min. The column was washed three to fivetimes between each sample with distilled water. Theeffluent was recorded by a flame ionization detector at140 C with hydrogen at 40 ml/min and air at 400ml/min. The signal was amplified to give a range ofzero to 10-9 amp. Volatile fatty acid peaks wereidentified by comparing their retention times withthose obtained with a control solution of 10 meq ofC2-C, fatty acids per liter of FEP. The height of asample peak was recorded as a percentage of theheight of the corresponding control peak. Alcoholpeaks, identified by comparison with a control solu-tion in FEP, were recorded as present or absent. Todetermine the presence of lactate and succinate, 1 mlof the acidified supernatant fluid was methylatedwith 1 ml of BF,-methanol (Applied Science) over-night at room temperature and extracted with 0.5 mlof chloroform. Two microliters of the chloroformextract was analyzed by using the same system asabove and a control solution containing 10 meq oflactate and succinate per liter. For the final analysis,the acid end products were recorded on a 1 to 4+system on the basis of the relative amounts in aculture as well as the absolute amounts.The results of the above tests were coded as the

presence or absence of 60 characteristics listed inTable 2; 25 of these characteristics are partiallydependent on other listed characteristics. For a veryfew isolates there were no data on one or twocharacters-for example, some strongly aerogenic iso-lates repeatedly disrupted the agar deeps.

One aim of the characterization of the isolates bythe method described above was to make it possible togroup together all bacteria of one strain, that is, ofrecent common ancestry, and to separate differentstrains even when they might belong to the samespecies. An automatic clustering program, modifiedfrom Quadling (13), with subroutine LINK (22) forsingle-link hierarchical clustering, run on the IBM360/75, was used for initial sorting of the results for allthe isolates from each host separately. Grouping wasbased mainly on the clusters formed by using the 6%coefficient of difference; that is, members of onecluster would differ from members of every othercluster by at least 4 out of 60 characters. However,clusters formed at other levels and taxonomic criteriafrom Holdeman and Moore's classification (11) werealso taken into account in forming the groups. Thiswas done to avoid both groups which were separatedby characters that might be caused by technicalvariation or changes in a single protein, and groupswhich included members that differed in charactersconsidered critical by taxonomists. The characters ofthe groups obtained were compared with the charac-ters given for species listed by Holdeman and Moore.

Detection of bacteria with fluorescent antibody.Antisera were prepared against two or more isolates ofgram-positive bacteria from each host by the follow-ing method. The bacteria were grown in AB at 35 C,with or without shaking, to maximal turbidity, har-vested by centrifugation at 4,000 x g for 30 min,washed three times in half-strength fluorescent trep-onemal antigen buffer (FTA; BBL) with 0.5% For-

TABLE 2. Morphological and cultural characteristicsof isolates listed as present, absent, or no data for

automatic clustering

No. Chararteristic

2

3

45

6

7

8

9

101112131415-1819-2223-2627-30

313233-3435-36373839404142434445-4849-5253-5657-60

Colonies grew in one day on initial subculture.Round coccal forms in Gram stain from agar

or broth.Oval coccal forms in Gram stain from agar or

broth.Rod forms in Gram stain from agar or broth.Filamentous forms in Gram stain from agar or

broth.Branching forms in Gram stain from agar or

broth.Chains of bacteria in Gram stain from agar or

broth.Clumps of bacteria in Gram stain from agar or

broth.Anaerobe, only grew more than 2 mm below

surface of PYS agar deep.Reduced resorufin in PYS agar deep.Produced acid from glucose.Produced marked acid from glucose in 2 days.Produced gas from glucose.Produced marked gas from glucose in 2 days.As 11 to 14 for maltose in place of glucose.As 11 to 14 for lactose in place of glucose.As 11 to 14 for sucrose in place of glucose.As 11 to 14 for soluble starch in place of

glucose.Produced acid from mannose.Produced gas from mannose.As 31 to 32 for xylose in place of mannose.As 31 to 32 for inulin in place of mannose.Produced acid from mannitol.Produced acid from inositol.Produced ethanol in FEP.Produced butanol in FEP.Produced 1 + to 4+ acetate in FEP.Produced 2+ to 4+ acetate in FEP.Produced 3+ to 4+ acetate in FEP.Produced 4+ acetate in FEP.As 41 to 44 for propionate in place of acetate.As 41 to 44, butyrate in place of acetate.As 41 to 44, lactate in place of acetate.As 41 to 44 for succinate in place of acetate.

malin, and resuspended in this buffer to match theturbidity of McFarland tube no. 4. Young albinorabbits were injected intravenously with 1 ml twice aweek for 4 weeks and for another week after aninterval of 3 weeks. Seven to 10 days after the lastinjection, the rabbit was bled and the serum washarvested.The globulin was separated (3, 10) and the concen-

tration was estimated by using a refractometer; ifnecessary the solution was diluted to a protein con-centration of 20 mg/ml. The solution was brought topH 9 with 10 x concentrated pH 9 buffer (4), andfluorescein isothiocyanate (FITC), isomer I (BBL),was added to give 25 sg of FITC to 1 mg of protein.

VOL. 9, 1974 721

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GOSSLING AND SLACK

The mixture was stirred thoroughly and refrigeratedovernight. Excess FITC was removed by dialysisagainst FTA buffer. The conjugate was filtered andstored at - 20 C.

Smears of pure cultures were stained with un-diluted FITC conjugate for 30 min at room tempera-ture, washed twice with FTA buffer and twice withpH 9.18 buffer (Fisher Scientific), and mounted inglycerol with pH 9 buffer. The slides were examinedwith a x95 objective on a Leitz Ortholux microscopeequipped for use with fluorescein stains. As a partialestimate of specificity, the FITC conjugates weretested against all isolates from the same specimen(including many of those which were lost before theywere fully characterized), and they were tested forcross-reactions with other isolates of the same type. Asatisfactory conjugate gave an even stain of the cellwalls of the isolate used in its preparation, and it alsostained isolates of the same group from the samespecimen. However, it did not stain dissimilar isolatesor isolates from other hosts even if they were of thesame cultural type. The conjugates were then testedagainst all isolates from the same host.The FITC conjugates were also used to stain the

smears prepared from the 10-3 dilution of feces fromthe corresponding hosts following a blocking tech-nique to increase specificity. To prepare the blockingsera, 1 ml of each antiserum, from the same lots usedto prepare the conjugates, was sorbed with 0.1 ml ofpacked cells of the antigen used in its production at45 C for 2 h, with occasional shaking, and in therefrigerator overnight. The cells were removed bycentrifugation at 800 x g for 30 min. This sorptionwas repeated twice and checked by staining a smear ofthe corresponding strain with the sorbed serum 30min at room temperature, washing with FTA, stain-ing for 30 min with goat anti-rabbit globulin 1:8(Microbiological Associates), washing, and mountingas above. The sorption was repeated until there wasno cell wall fluorescence.The fecal smears were stained at room tempera-

ture. They were covered with a drop of the sorbedantiserum for 30 min. The antiserum was removedwith a capillary pipette applied to one corner; thiscorner was not used for counting. A drop of thespecific FITC conjugate containing rhodamine-conju-gated bovine serum albumin (NBCo) as a counter-stain (the exact amount was adjusted empirically foreach conjugate) was then placed on the smear for 30min and withdrawn as before. The smears were

washed three times with FTA buffer and twice withpH 9.18 buffer by dropping the buffer and withdraw-ing it as above, and then were mounted in bufferedglycerol.The fluorescing bacteria were counted in 50 fields

in each smear in the same manner as used for the totalmicroscope counts. The total bacteria were counted infive fields in each smear by using white light anddarkfield illumination. If the count was less than 80%of the original total count, the results were discardedand another slide was stained. The fluorescent bacte-ria counted were multiplied by 8.414 x 101 to give thecount of such bacteria per gram (wet weight).

RESULTSThe direct microscopic counts and total cul-

tural counts were calculated on the basis of wetweight of specimen, dry weight of specimen,and dry weight of nonmicrobial residue (Table3). The counts estimated on each basis wereequally variable. Therefore the wet weight ba-sis, which requires the minimum of calculation,was chosen for all further analyses.Microscopic counts. The 47 specimens aver-

aged 1011.1 organisms/g (wet weight) as esti-mated by direct microscopic counts (Table 3).By using the Gram stain, gram-positive rodsaccounted for 21% of the organisms seen, andgram-positive cocci accounted for 5% (Table 4).There was a positive correlation between thecounts of gram-positive cocci and gram-positiverods (correlation coefficient = 0.76; P = 0.0001).Altogether, gram-positive bacteria accountedfor 25% of organisms seen, varying from 2 to 45%in individual specimens.Cultural counts. The total counts on RF agar

are given in Table 3, and the estimated differen-

TABLE 3. Comparison of total bacterial countscalculated on different bases

Geo- Coeffi-Type of Basis metric SD ocfientcount mean (log10) iation

(10910) ~ (%)

Microscopic Per g (wet wt) 11.13 0.25 2.3Per g (dry wt) 11.66 0.23 1.9Per g (dry residue) 11.70 0.25 2.1

Cultural Per g (wet wt) 10.86 0.30 2.8Per g (dry wt) 11.39 0.31 2.7Per g (dry residue) 11.44 0.31 2.7

aTotal of 47 specimens from 10 hosts: 3 from oneand 4 from another and 5 from each of the remaining8.

TABLE 4. Estimated counts of bacteria differentiatedby Gram reaction and microscopic morphology

Microscopic counts Cultural countsGram Mor- (log10) (log10)reac- phol- Geo- Geo-tion ogy metric SD Range metric SD Range

meana mean

+ Rods 10.37 0.38 9.3-10.9 10.26 0.40 9.1-11.1+ Cocci 9.750 0.42 8.5-10.6 10.04 0.50 9.0-11.0_ Rods 10.93 0.28 10.3-11.4 10.47 0.27 9.9-11.1- Cocci 9.88 0.47 8.5-10.5 9.6 10.49 8.6-10.6

'Total of 47 fecal specimens from 10 hosts: 3 from one, 4from another, and 5 from each of the remaining 8.

' Only seen in 46 specimens.'Only seen in 40 specimens.

722 INFECT. IMMUNITY

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GRAM-POSITIVE BACTERIA IN HUMAN FECES

tial counts are given in Table 4. The totalcounts of bacteria cultured averaged 55% (for 46specimens, with a range of 8 to 175%) of thoseseen on direct microscopic examination. Differ-ential counts showed higher proportions ofgram-positive bacteria and of cocci than thedirect microscopic examination: 29% of thecolonies were of gram-positive rods and 19%were of gram-positive cocci. The mean apparentrecovery rates for the different groups were asfollows: gram-positive rods, 128%; gram-posi-tive cocci, 334%; gram-negative rods, 47%; andgram-negative cocci, 93%. Altogether, 47% ofthe bacteria grown were gram positive; theyvaried from 16 to 84% in individual specimens.The host was a significant source of variation inthe number of gram-positive bacteria cultured(P = 0.02).

Isolation and characterization. Seventy-two percent of the colonies picked for isolationgrew on subculture; the recovery rate variedgreatly between specimens, ranging from 55 to98% for the 43 specimens from which 50 or morecolonies were picked. Fifty-seven percent of theisolates stained gram positive; this percentagevaried from 25 to 85% for individual specimens.The proportion of bacteria picked which grewand were gram positive was nearly the same asthe proportion of colonies on RF agar whichwere gram positive, indicating that few gram-positive bacteria were lost on subculturing.However, 24% of these gram-positive bacteriawere lost on subsequent processing, leaving 865isolates which were characterized (Table 5).

TABLE 5. Bacterial isolates subcultured,characterized, and grouped on the basis of cultural

characteristicsNo. of

No. of bacterial isolates groupsNo. of distin-speci- guished

Host mens Gram on basiscul- Sub- Grew + on Char- of cul-tured cul- on sub- sub- acter- tural

tured culture culture ized charac-teristics

I 3 185 149 97 78 12II 5 268 174 94 55 24

III 5 432 281 91 58 28IV 5 300 235 162 130 34V 4 184 147 72 53 19VI 5 306 251 173 152 27VII 5 245 163 112 96 23VIII 5 295 214 95 82 39IX 5 295 212 136 103 27X 5 263 184 104 58 26

Total 47 2,773 2,018 1,137 865

The majority of the tests used in characteriz-ing the complete set of isolates examined carbo-hydrate fermentation (fermentation productsand range of carbohydrates attacked) because apreliminary survey of 92 isolates had shown that67 of them were more active in carbohydratemedia than in media containing peptone, lac-tate, or lipid as an energy source, and that 86 ofthem produced acid from sugars in 7 days.

After characterization, it was apparent thatthere was a large variety of bacteria, evenamong isolates from a single specimen. Thenumber of groups of isolates distinguished ineach host is given in Table 5. The averagenumber was 28.5 (range 24 to 39) per host for theeight hosts from which five specimens wereexamined. The average number of groups dis-tinguished from individual specimens was 10.0(range 3 to 15) per specimen for the 25 speci-mens from which 15 or more isolates werecharacterized. Many of the groups could betentatively identified with species described byHoldeman and Moore (11). All groups thoughtto belong to a single species are referred totogether as a type. Thus, by the definition givenunder methods, isolates in the same group comefrom the same host and are thought to belong tothe same strain, but isolates of the same typemay come from different hosts and are thoughtto belong to the same species. The types foundin more than one individual are listed in Table6. The characters of these types are those listedby Holdeman and Moore (11) for the speciesindicated, but the types were not definitelyidentified because the tests used did not includeall the definitive tests for these genera andspecies. Each type was readily distinguishablefrom the others except that type 3 (resemblingEubacterium contortum) was close to somegroups of type 9 (resembling Peptostreptococ-cus productus). Seventy-four percent of theisolates belonged to groups which could betentatively identified, and 60% were tentativelyidentified as Bifidobacterium adolescentis, Eu-bacterium aerofaciens, Eubacterium rectale, P.productus, or Ruminococcus bromii.The isolates of type 1 are all bifidobacteria but

might include some Bifidobacterium longum aswell as B. adolescentis. The isolates resemblingE. aerofaciens are not all clearly distinguishedfrom Clostridium ramosum, but most of theisolates resembled E. aerofaciens more closelybecause of coccal forms, anaerogensis, or failureto ferment amygdalin. The isolates resemblingE. rectale might include some clostridia. Theisolates resembling P. productus had a charac-teristic morphology but were heterogeneous in

723VOL. 9, 1974

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GOSSLING AND SLACK

TABLE 6. Occurrence and tentative identification ofpredominant types of isolates based on cultural

characteristics

Found in:

Tnaie No. of % ofType Tentative ig so- No. of % of No.of

latesa lates speci- speci- hostsdmens b mens' o

1 Bifidobacterium 84 10 22 50 9adolescentis

2 Eubacterium 160 18 28 68 8aerofaciens

3 E. contortum 14 2 5 13 44 E. cylindroides 36 4 10 26 65 E. rectale 80 9 24 58 106 E. tortuosum 7 1 4 11 37 E. ventriosum 13 1 9 21 58 Propionibacte- 10 1 3 8 3

rium acnes9 Peptostreptoc- 77 9 27 68 9

cus productus10 Ruminococcus 9 1 5 11 3

albus11 R. bromii 115 13 21 55 1012 Sarcina 12 1 2 5 2

ventriculi

aTotal isolates-865.'Total specimens-47.c Total specimens (excluding those from which less than 9

isolates were characterized)-38.d Total hosts-10.

fermentative characters. More than one groupfrom some specimens resembled P. productus asdescribed by Holdeman and Moore (11). Theisolates resembling R. bromii were quite dis-tinctive, and representative isolates from eachhost all stained with a single FITC conjugate.Estimated geometric mean counts of these

five types were as follows: bifidobacteria, 109 5/g(wet weight); E. aerofaciens-like, 1097/g (wetweight); E. rectale-like, 1094/g (wet weight); P.productus-like, 1094/g (wet weight); and R.bromii-like 1097/g (wet weight). Analysis ofvariance of the counts by host and by type ofbacteria, and the interaction of these two fac-tors, indicated that the counts varied with thehost and the host-type interaction but did notvary with the bacterial type (Table 7). Thehost-type interaction was obvious with the E.aerofaciens type, which was absent or rare inthree hosts and common in seven. Strains ofthis type, distinguished by FITC conjugates,accounted for 5 to 6% of the total direct count inspecimens I/I, VI/4, and X/3, a much higherproportion than normally found (compare Ta-bles 3 and 9).

Further analysis of the estimated countssuggested that the counts of bifidobacteria(type 1), E. rectale-like (type 5), R. bromii-like(type 11), and gram-negative bacteria wereintercorrelated. Correlation coefficients (P <

0.05) for the 43 specimens from which 50 ormore colonies were picked were as follows:between 1 and 5, 0.72; between 1 and 11, 0.73;between 1 and gram-negative rods, 0.56; be-tween 1 and gram-negative cocci, 0.72; between5 and 11, 0.67; between 5 and gram-negativerods, 0.64; and between 5 and gram-negativecocci, 0.70. Also, estimated counts of type 4,which resembled E. cylindroides, and types 9and 11 were negatively correlated with theresidual water content of the feces, obtained bysubtracting the dry weight and the estimatedwater content of the bacteria seen in directsmear (17) from the wet weight. The correlationcoefficients (P < 0.05) were as follows: forresidual water with 4, -0.67; with 9, -0.42; andwith 11, -0.65.The 29% of isolates not included in Table 6

were mostly grouped in ones or twos and did notobviously resemble isolates from other hosts. Itis certain that there was no unidentified typecomparable in importance with the five de-scribed above.Persistence of bacterial strains as indi-

cated by grouping bacteria on cultural char-acters. Within individual hosts, many groups ofbacteria were found to be present in successivespecimens. When there were two consecutivespecimens from the same host from which 15 ormore isolates had been characterized, an aver-age of 35% (range 14 to 55) of the isolates fromthe second specimen were grouped with isolatesfrom the first. From the eight hosts from whichfive specimens were collected, an average of 46%(range 21 to 80) of the isolates from the finalspecimen were grouped with isolates from previ-ous specimens. The persistence of individualgroups is listed in Table 8.Persistence of bacterial strains as indi-

cated by serological grouping of isolates anddirect staining of fecal smears with FITC

TABLE 7. Analysis of variance of estimated culturalcounts of five types ofgram-positive bacteria from 43fecal specimens from which 50 or more colonies were

isolated

Degrees of ProbabilitvSource of variation freedom of the null

hypothesisa

Host 9 0.0022Type of bacterial 4 0.2126Interaction between host 36 0.0176and type of bacteria

Error 185

a That the given source of variation did not influ-ence the count obtained.

'Types resembling B. adolescentis, E. aerofaciens,E. rectale, P. productus, and R. bromii.

INFECT. IMMUNITY724

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GRAM-POSITIVE BACTERIA IN HUMAN FECES

TABLE 8. Persistence of culturally defined groups ofgram-positive bacteria in individual hosts

No. of No. of Days be-groups these awend larst

No. of No. of distin- groups and last

Host speci- isolates guished found in specimenmens charac- on cul- more memberscocultured terized tural than one the same

charac- speci- gru weteristics men isolated

I 3 78 12 1 281II 5 55 24 5 59, 344, 419,

578, 578III 5 58 28 5 51, 380, 480,

580, 580IV 5 130 34 11 77, 77, 147,

147, 253,343, 343,449, 449,519, 596

V 4 53 19 3 70, 84, 84VI 5 152 27 10 98, 314, 446,

446, 446,446, 591,591, 591,591

VII 5 96 23 7 84, 238, 345,350, 350,588, 588

VIII 5 82 39 8 116, 237,237, 237,237, 237,311, 548

IX 5 103 27 8 135, 238,382, 485,485, 620,620, 620

X 5 58 26 4 222, 510,510, 638

conjugates. Eighteen FITC conjugates distin-guished between isolates of the same culturaltype, that is, isolates resembling a single spe-

cies, and were therefore used for serologicalgrouping. Six conjugates were prepared byusing isolates of bifidobacteria; there was no

cross-reaction between these conjugate isolatesystems, nor were these isolates stained byconjugates prepared with ATCC B. adolescentisstrains 15703, 15704, 15705, and 15706. Thus,these conjugates defined six serological groupsof bifidobacteria. In fecal smears these conju-gates stained large irregular rods, but actualbifid forms were rarely seen; only one bifid formin 500 cells was observed in one specimen wherethis was recorded.Seven conjugates were prepared by using E.

aerofaciens type isolates; two of these conjugateisolate systems cross-reacted. Thus, six serologi-cal groups were distinguished, one of which was

found in two hosts. On fecal smears they stainedsmall rods of regular diameter, cocci, andoccasional chains and filaments up to 20 gm

long. A conjugate prepared by using an E.cylindroides-type isolate and tested againstsuch isolates from five other hosts stained oneisolate from another host. On fecal smears itstained long regular rods of constant diameter.Three conjugates were prepared by using P.productus-type isolates; there were no cross-reactions, and thus three serological groupswere defined. On fecal smears these conjugatesstained large oval or pointed cocci, often inpairs, and sometimes elongated into cigar-shaped rods. A conjugate prepared by using oneunidentified isolate was tested against similarisolates from two other hosts and stained one ofthem well. On fecal smears it stained smallirregular rods.

Fluorescein isothiocyanate conjugates pre-pared by using R. bromii-type isolates andPropionibacterium acnes-type isolates were ofno use for distinguishing individual strainsbecause they stained all isolates of the sametype. Nevertheless, fecal smears from one host(I) were stained with a conjugate prepared byusing a P. acnes-type isolate in order to verifythat propionibacteria were actually present inthe feces; small rods in the fecal smears werestained (Table 9).The results of staining the isolates and direct

fecal smears from each host with FITC conju-gates prepared by using isolates from the samehost are shown in Table 9. The correspondencebetween the counts estimated by staining directsmears and those estimated by staining isolatesis not good, but both methods showed persist-ence of serologically defined groups. One groupresembling B. adolescentis, five resembling E.aerofaciens, and three resembling P. productuswere present for more than 18 months. Thesimilarity of counts on direct smears from asingle specimen made by using different conju-gates might suggest that the same nonspecifi-cally stained organisms were counted repeat-edly, but this was not so since the differenttypes had distinct morphologies as describedabove.

DISCUSSIONIn this study the gram-positive bacteria cul-

tured averaged 10105/g (wet weight) of feces,which is more than that reported in mostsurveys of fecal flora but the same as thatreported by Drasar et al. (5), who also used'astrict anaerobic technique. In determining theproportion of the fecal flora that was grampositive, the figure indicated by direct examina-tion (average 25%) was probably too low, sincesome individual gram-positive bacteria maystain gram negative. The figure (average 47%)

VOL. 9, 1974 725

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

TABLE 9. Persistence of serologically defined groups of gram-positive bacteria in individual hosts

Log,0 of Log,0 ofDay of count Day of count

Tentative iden- specimen Log,0 of estimated Tentative iden- specimen Log10 of estimatedtification of collection count from pro- tification of collection count from pro-

isolate used in as counted from di- portion of isolate used in as counted from di- portion ofHost the preparation from first rect smear isolates Host the preparation from first rect smear isolatesof FITC collection with which of FITC collection with whichwith withconjugate from the stained conjugate from the ugate stained

host listed conjugate with host listed conj withconjugate conjugate

I

II

III

IV

V

VI

Eubacteriumaerofaciens

Propionibacte-rium acnes5

Eubacteriumaerofaciense

Eubacteriumaerofaciense

Peptostreptococ-cus productus

Eubacteriumcylindroides

Bifidobacteriumadolescentis'

Bifidobacteriumadolescentis'

Eubacteriumaerofaciens

Peptostreptococ-cus productus

0112281

0112281

0

59293419637

0

59293419637

0

100329380580

0

147253519596

0

84328329

0

84328398

0

145216314591

0

145216314591

9.89.28.5

8.08.3c

_d

8.5

8.04.

9.39.3

9.09.79.7

9.09.19.58.38.0

8.7

8.68.7

9.39.39.510.09.7

9.29.59.39.89.8

10.79.39.2

-a10.2'

10.1

9.7

9.89.5

8.99.8

10.2

10.99.3

9.39.1

9.99.910.610.110.1

9.1

9.610.2

VII

VIII

Ix

x

Bifidobacteriumadolescentis

Eubacteriumaerofaciens

Peptostreptococ-cus productus

Bifidobacteriumadolescentis

Bifidobacteriumadolescentis

Eubacteriumaerofaciens

Unidentifiedbacillus

Bifidobacteriumadolescentis

Eubacteriumaerofaciens

084

322429672

084

322429672

084

322429672

0116220311548

0135266382620

0135266382620

0135266382620

0128303416638

0128303416638

9.49.69.19.59.3

9.19.19.29.69.1

9.19.49.19.49.1

9.1

10.18.29.0

i

9.7

9.79.310.09.68.6

a No isolates stained.'With the techniques used in this study, all P. acnes stain with one conjugate.'Between 2 and 10 stained bacteria in 100 fields, representing 107 to 108/g (wet weight).d Zero or one stained bacteria in 100 fields, representing less than 107/g (wet weight).eTwo serologically distinct groups.' Two serologically distinct groups.

726

10.7

9.1

9.99.19.4

9.9

9.8

9.810.1

9.6

9.99.69.4

10.6

10.1

9.310.0

9.99.39.3

9.69.910.29.39.5

-1

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GRAM-POSITIVE BACTERIA IN HUMAN FECES

indicated by the cultural counts was probablytoo high, since it appeared that many of thegram-negative bacteria did not grow on culture.It is estimated that the true proportion isprobably between 10 and 50% in most speci-mens, with the host being a significant factor indetermining the actual proportion. It is to benoted that in this investigation the host variableincludes diet and other habits which were notcontrolled.The gram-positive bacteria isolated were all

anaerobes and fermented carbohydrates, butotherwise they were a heterogeneous group notreadily represented by a single metabolic typesuch as bifidobacteria or any one species ofEubacterium. This heterogeneity has been pre-viously suggested by others studying smallergroups of subjects (1, 16).The variety of types found in individual

specimens meant that any one type accountedfor only a small proportion of the flora. Thecount of individual strains distinguished byFITC conjugates on direct smears was usuallyabout 1% of the total count. The averageestimated cultural count of the five major typeswas from 2 to 5%, indicating cultural enrich-ment. Since the sampling procedures were de-signed on the assumption that predominantstrains would constitute 20% of the flora, thedata obtained were less useful than was an-ticipated. Nevertheless, it was shown that therewere significant host idiosyncrasies in the com-position of the gram-positive fecal flora, andsome interrelationships were suggested. It wasinteresting that the bifidobacteria counts cor-related with those of gram-negative anaerobessince bifidobacteria have been considered analternative to a putrefactive or gram-negativeflora (8).

Five types of bacteria were distinguished onthe basis of cultural characteristics which ac-counted for 60% of the bacteria isolated; eachtype resembles a species which has been re-ported previously from human feces (11). Threeof these types, resembling B. adolescentis, E.rectale, and P. productus, ferment a wide rangeof carbohydrates and in the gut probably makeuse of carbohydrates that the host cannot digestor absorb. Another type, resembling R. bromii,fermented primarily maltose but also solublestarch, glucose, and sometimes mannose; in thegut it probably makes use of starch which wasnot digested fast enough for the products to beabsorbed by the host. The other type, resem-bling E. aerofaciens, ferments several carbohy-drates but does not fall clearly into either of theabove categories. This type was most responsi-ble for the variation between hosts. In a few

specimens it accounted for a relatively highproportion of the flora.A major purpose of this study was to distin-

guish the predominant strains or populations ofthe gram-positive bacteria within each host andto determine whether they were resident ortransient. Two methods were used to distin-guish bacteria representing particular strains.The first method, grouping of isolates on theresults of cultural tests, presented problems inthe selection of tests, the recording of results,and the method of clustering. Only a shortseries of tests could be handled and, afterpreliminary tests, tests were selected whichwere reproducible and distinguished betweenisolates. This series was inadequate for taxo-nomic purposes, and therefore subsequent iden-tification of the groups was only tentative. Inrecording the results of the tests an attempt wasmade to record some of the features that areused in visual recognition of strains when smallnumbers of isolates are examined. Such featureswere the pattern of acid and gas production andthe pattern of peaks produced by chromatogra-phy of the fermentation end products. In bothrecording and clustering, it is necessary to useuniform methods for all isolates, but thesemethods may produce distortion when appliedto a wide range of organisms because some typesshow more phenotypic variation than others. Inclustering, the single-link method is sensitive tothe numbers in a potential cluster. Large clus-ters will commonly be recognized at a lowerlevel of difference than small ones. Together,these problems resulted in some anomalousclusters which were edited by using taxonomiccriteria from more extensive studies (11). Inevery host this editing resulted in fewer groupsthan there were clusters at the 6% level ofdifference.A particular problem of the second method

for detecting strains, the use of FITC conju-gates, was in determining the specificity of theconjugates. Strain-specific conjugates should.not cross-react with unrelated bacteria andshould distinguish between strains of a singlespecies. Testing for cross-reactions with un-related bacteria has always been a problemwhen fluorescent antibody techniques are usedin the direct examination of feces since there areso many unknown bacteria in feces. Neverthe-less, such techiques are used in the detection ofpathogens. In this study three factors suggestedthat the organisms stained were homologouswith those used in preparing the conjugate.First, the conjugates used did not cross-reactwith a sample of gram-positive bacteria isolatedfrom the same specimen as the isolate used in

VOL. 9, 1974 727

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GOSSLING AND SLACK

preparing the conjugate. Secondly, the orga-nism stained had a characteristic morphology.Thirdly, significant numbers of stained orga-nisms were only seen in fecal smears from hostsfrom which bacteria of the same serological andcultural type had been isolated. Even if conju-gates do not cross-react with unrelated bacteria,they are not necessarily suitable for tracingstrains. Serological methods have been used totrace strains of E. coli in the fecal flora (2, 19,23), but E. coli was already known to beserologically diverse. The bacteria isolated here,apart from bifidobacteria (18) and propionibac-teria (12), have not been previously studiedserologically. Therefore the conjugates weretested against isolates of the same type fromother hosts. Such tests indicated that the bi-fidobacteria and the isolates resembling E.aerofaciens could be divided into several sero-logical groups so the conjugates would be usefulin tracing strains, that strains of R. bromii andP. acnes could definitely not be distinguishedby these conjugates, and that some other types,E. cylindroides-like, P. productus-like, and anunidentified bacillus, could at least be subdi-vided by using the conjugates; but there weretoo few conjugates to establish the degree ofheterogeneity.Grouping on the basis of cultural characteris-

tics allowed all isolates to be grouped. Serologi-cal grouping is more definitive but could onlygroup a few isolates (22%). The serologicalgroups correspond largely with cultural groupsexcept that one cultural group of bifidobacteriaencompassed two serological groups.The cultural grouping suggested that there

were resident strains in each host and that,allowing for sampling deficiencies, probably atleast half of the strains were resident. When theFITC conjugates were used to detect strains, itwas found that E. aerofaciens-like bacteria weregenerally represented by a single strain whichwas a long-term resident. Resident strains ofbifidobacteria were also found, but they ap-peared to change more frequently than the E.aerofaciens-like strains. Resident strains of P.productus and an unidentified bacillus weresuggested.This technique of staining fecal smears with

FITC conjugates for species or strains of bacte-ria might find useful application as a quickmethod of obtaining counts in dietary andmedical surveys. In most cases the staining wasvery sharp and specific and would have beensuitable for automated counting.

In conclusion, gram-positive bacteria form alarge portion of the human fecal flora. Althoughthey all ferment carbohydrate, they cannot be

considered as a metabolic unit in studyingcolonic ecology since they belong to variousphysiological types, each with a distinctivemetabolism, and the relative numbers of thesetypes vary from host to host. There is alsovariation within each host over time, but somestability of the flora is indicated by both theidiosyncratic patterns of counts detected overthe 18-month period and by the persistence ofsome serologically defined strains over the sameperiod.

ACKNOWLEDGMENTS

This investigation was supported in part by the PublicHealth Service general research support grants 5S01RR05433-09 and -10.We are very grateful to Mary Ann Gerencser for many

helpful suggestions.

LITERATURE CITED

1. Attebery, H. R., V. L. Sutter, and S. M. Finegold. 1972.Effect of a partially chemically defined diet on normalhuman fecal flora. Amer. J. Clin. Nutr. 25:1391-1398.

2. Branche, W. C., Jr., V. M. Young, H. G. Robinet, and E.D. Massev. 1963. Effect of colicine production onEscherichia coli in the normal human intestine. Proc.Soc. Exp. Biol. Med. 114:198-201.

3. Burrell, R. G., and C. C. Mascoli. 1970. Experimentalimmunology, 3rd ed. Burgess Publishing Co., Min-neapolis.

4. Cherry, W. B., M. Goldman, T. R. Carski, and M. D.Moody. 1960. Fluorescent antibody techniques in thediagnosis of communicable diseases. Public HealthService Publications no. 729, Washington, D.C.

5. Drasar, B. S., M. Shiner, and G. M. McLeod. 1969.Studies on the intestinal flora. I. The bacterial flora ofthe gastrointestinal tract in healthy and achlorhydricpersons. Gastroenterology 56:71-79.

6. Emslie-Smith, A. H. 1965. Observations on strains ofEscherichia coli exhibiting ecological dynasticism inthe fecal flora of man. Ernahrungsforsch. 10:302-307.

7. Gorbach, S. L., L. Nahas, P. I. Lerner, and L. Weinstein.1967. Studies of intestinal microflora. I. Effects of diet,age and periodic sampling on numbers of fecal microor-ganisms in man. Gastroenterology 53:845-855.

8. Haenel, H. 1970. Human normal and abnormal gastroin-testinal flora. Amer. J. Clin. Nutr. 23:1433-1439.

9. Haenel, H., B. Gassman, F. K. Griutte, and W. Muller-Beuthow. 1964. Einflusse einer zellulosereichen Kostauf die intestinale Mikro6kologie beim Menschen.Zentralbl. Bakteriol. Parasitenk. Abt. I Orig.192:491-499.

10. Hebert, G. A., B. Pittman, R. M. McKinney, and W. B.Cherry. 1972. The preparation and physicochemicalcharacterization of fluorescent antibody reagents. U.S.Department of Health, Education, and Welfare, PublicHealth Services and Mental Health Administration,Center for Disease Control, Laboratory Division, At-lanta, Ga.

11. Holdeman, L. V., and W. E. C. Moore (ed.). 1972.Anaerobe laboratory manual. Virginia Polytechnic In-stitute Anaerobe Laboratory, Virginia Polytechnic In-stitute and State University, Blacksburg, Va.

12. Holmberg, K., and U. Forsum. Identification of Ac-tinomyces. Arachnia, Bacterionema, Rothia, and Pro-pionibacterium species by defined immunofluores-c.ence. Appl. Microbiol. 25:834-843.

13. Lockhart, W. R., and J. Liston (ed.). 1970. Methods for

728 INFECT. IMMUNITY

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from

GRAM-POSITIVE BACTERIA IN HUMAN FECES

numerical taxonomy. American Society for Microbiol-ogy, Bethesda, Md.

14. Mata, L. J., C. Carrillo, and E. Villatoro. 1969. Fecalmicroflora in healthy persons in a preindustrial region.Appl. Microbiol. 17:596-602.

15. Mitsuoka, T., T. Sega, and S. Yamamoto. 1965. Eineverbesserte Methodik der qualitativen und quan-

titativen Analyse der Darmflora von Menschen undTieren. Zentralbl. Bakteriol. Parasitenk. Abt. I Orig.195:455-469.

16. Moore, W. E. C., and L. V. Holdeman. 1972. Identifica-tion of anaerobic bacteria. Amer. J. Clin. Nutr.25:1306-1313.

17. Paigen, K., and B. Williams. 1970. Catabolite repressionand other control mechanisms in carbohydrate utiliza-tion. Advan. Microbial Physiol. 4:251-324.

18. Reuter, G. 1963. Vergleichende Untersuchung iuber dieBifidus-flora im Sauglings und Erwachsenenstuhl.Zentralbl. Bakteriol. Parasitenk. Abt. I Orig.191:486-507.

19. Sears, H. J., H. Janes, R. Saloum, I. Brownlee, and L. F.Lamoureaux. 1956. Persistence of individual strains ofEscherichia coli in man and dog under varying condi-tions. J. Bacteriol. 71:370-372.

20. Seeliger, H., and H. Werner. 1963. Recherches qualita-tives et quantitatives sur la flore intestinale del'homme. Ann. Inst. Pasteur 105:911-936.

21. Service, J. 1972. A user's guide to the statistical analysissystem. Student Supply Stores, North Carolina StateUniversity, Raleigh, N.C.

22. van Rijsbergen, C. J. 1970. A fast hierarchic clusteringalgorithm. Computer J. 13:324-326.

23. Wiedemann, B., and H. Knothe. 1969. Untersuchungeniiber die Stabilitat der Koliflora des gesunden Mensc-hen 1 Mitteilung. Uber das Vorkommen permanentenund passanten Typen. Arch. Hyg. Bakteriol.153:342-348.

24. Zubrzycki, L., and E. H. Spaulding. 1962. Studies on

the stability of the normal human fecal flora. J.Bacteriol. 83:968-974.

VOL 9, 1974 729

on May 15, 2020 by guest

http://iai.asm.org/

Dow

nloaded from